{
  "id": 74765,
  "title": "How to train locally and push the model to kaggle for evaluation?",
  "url": "/competitions/humpback-whale-identification/discussion/74765",
  "author_name": "",
  "post_date": "2018-12-15T07:36:47.877970400Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Hi,\nI have a model which takes more than the 6hrs tier limit provided by kaggle to train. I an train the model in my local computer in few days and get the predictions. What is the procedure to submit these local predictions for evaluation in kaggle?\nIs it mandatory that the models should be trained via a kernel commit and the output of committed kernel can only be used?</p>",
  "messages": [
    {
      "id": "439337",
      "postDate": "12/15/2018 07:36:47",
      "content": "<p>Hi,\nI have a model which takes more than the 6hrs tier limit provided by kaggle to train. I an train the model in my local computer in few days and get the predictions. What is the procedure to submit these local predictions for evaluation in kaggle?\nIs it mandatory that the models should be trained via a kernel commit and the output of committed kernel can only be used?</p>",
      "rawMarkdown": "Hi,\nI have a model which takes more than the 6hrs tier limit provided by kaggle to train. I an train the model in my local computer in few days and get the predictions. What is the procedure to submit these local predictions for evaluation in kaggle?\nIs it mandatory that the models should be trained via a kernel commit and the output of committed kernel can only be used?",
      "votes": null
    },
    {
      "id": "439369",
      "postDate": "12/15/2018 09:10:25",
      "content": "<p>You can write the submission as a txt file, and then upload it.\nFor example:</p>\n\n<pre><code>fout = open('submission.csv','w')\nfout.write('Image,Id'+'\\n') \nfor name in all_name:\n    fout.write(name+',')\n    fout.write(prediction_0+' ')\n    fout.write(prediction_1+' ')\n    fout.write(prediction_2+' ')\n    fout.write(prediction_3+' ')\n    fout.write(prediction_4+'\\n')\nfout.close()\n</code></pre>",
      "rawMarkdown": "You can write the submission as a txt file, and then upload it.\nFor example:\n\n    fout = open('submission.csv','w')\n    fout.write('Image,Id'+'\\n') \n    for name in all_name:\n        fout.write(name+',')\n        fout.write(prediction_0+' ')\n        fout.write(prediction_1+' ')\n        fout.write(prediction_2+' ')\n        fout.write(prediction_3+' ')\n        fout.write(prediction_4+'\\n')\n    fout.close()",
      "votes": null
    },
    {
      "id": "439433",
      "postDate": "12/15/2018 12:50:57",
      "content": "<p>Thanks :)\nNow only I noticed that option in the page.</p>",
      "rawMarkdown": "Thanks :)\nNow only I noticed that option in the page.",
      "votes": null
    },
    {
      "id": "440831",
      "postDate": "12/18/2018 02:25:18",
      "content": "<p>Ahh, I was assuming a CSV would be comma-separated, but it looks like the classes are space-separated, while the filename is comma-separated. Thanks.</p>\n\n<blockquote>\n  <p>Image,Id</p>\n  \n  <p>00028a005.jpg,new_whale w_23a388d w_9b5109b w_9c506f6 w_0369a5c</p>\n  \n  <p>000dcf7d8.jpg,new_whale w_23a388d w_9b5109b w_9c506f6 w_0369a5c</p>\n  \n  <p>000e7c7df.jpg,new_whale w_23a388d w_9b5109b w_9c506f6 w_0369a5c</p>\n  \n  <p>...</p>\n</blockquote>",
      "rawMarkdown": "Ahh, I was assuming a CSV would be comma-separated, but it looks like the classes are space-separated, while the filename is comma-separated. Thanks.\n\n&gt; Image,Id\n\n&gt; 00028a005.jpg,new_whale w_23a388d w_9b5109b w_9c506f6 w_0369a5c\n\n&gt; 000dcf7d8.jpg,new_whale w_23a388d w_9b5109b w_9c506f6 w_0369a5c\n\n&gt; 000e7c7df.jpg,new_whale w_23a388d w_9b5109b w_9c506f6 w_0369a5c\n\n&gt; ...",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 439369,
      "author_name": "luyang2018",
      "author_url": "",
      "post_date": "12/15/2018 09:10:25",
      "content": "<p>You can write the submission as a txt file, and then upload it.\nFor example:</p>\n\n<pre><code>fout = open('submission.csv','w')\nfout.write('Image,Id'+'\\n') \nfor name in all_name:\n    fout.write(name+',')\n    fout.write(prediction_0+' ')\n    fout.write(prediction_1+' ')\n    fout.write(prediction_2+' ')\n    fout.write(prediction_3+' ')\n    fout.write(prediction_4+'\\n')\nfout.close()\n</code></pre>",
      "votes": null,
      "replies": [
        {
          "id": 439433,
          "author_name": "deepiblis",
          "author_url": "",
          "post_date": "12/15/2018 12:50:57",
          "content": "<p>Thanks :)\nNow only I noticed that option in the page.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 440831,
          "author_name": "badtyprr",
          "author_url": "",
          "post_date": "12/18/2018 02:25:18",
          "content": "<p>Ahh, I was assuming a CSV would be comma-separated, but it looks like the classes are space-separated, while the filename is comma-separated. Thanks.</p>\n\n<blockquote>\n  <p>Image,Id</p>\n  \n  <p>00028a005.jpg,new_whale w_23a388d w_9b5109b w_9c506f6 w_0369a5c</p>\n  \n  <p>000dcf7d8.jpg,new_whale w_23a388d w_9b5109b w_9c506f6 w_0369a5c</p>\n  \n  <p>000e7c7df.jpg,new_whale w_23a388d w_9b5109b w_9c506f6 w_0369a5c</p>\n  \n  <p>...</p>\n</blockquote>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "439337": "Hi,\nI have a model which takes more than the 6hrs tier limit provided by kaggle to train. I an train the model in my local computer in few days and get the predictions. What is the procedure to submit these local predictions for evaluation in kaggle?\nIs it mandatory that the models should be trained via a kernel commit and the output of committed kernel can only be used?",
    "439369": "You can write the submission as a txt file, and then upload it.\nFor example:\n\n    fout = open('submission.csv','w')\n    fout.write('Image,Id'+'\\n') \n    for name in all_name:\n        fout.write(name+',')\n        fout.write(prediction_0+' ')\n        fout.write(prediction_1+' ')\n        fout.write(prediction_2+' ')\n        fout.write(prediction_3+' ')\n        fout.write(prediction_4+'\\n')\n    fout.close()",
    "439433": "Thanks :)\nNow only I noticed that option in the page.",
    "440831": "Ahh, I was assuming a CSV would be comma-separated, but it looks like the classes are space-separated, while the filename is comma-separated. Thanks.\n\n&gt; Image,Id\n\n&gt; 00028a005.jpg,new_whale w_23a388d w_9b5109b w_9c506f6 w_0369a5c\n\n&gt; 000dcf7d8.jpg,new_whale w_23a388d w_9b5109b w_9c506f6 w_0369a5c\n\n&gt; 000e7c7df.jpg,new_whale w_23a388d w_9b5109b w_9c506f6 w_0369a5c\n\n&gt; ..."
  },
  "source": "meta"
}